Fit curve to a model problem

Hi dear users,

I have a set of data in the graph below

And I'm trying to fit this data to a Magic formula tyre model which looks like this

Where D,C,B, and E are coefficients that I need to get from Matlab.

Now I've tried the fitting tool and inserting this custom model but I can never get it to fit even remotely good.

Anyone have any ideas?

Answer 1

[the cyclist]

Personally, I would use fitnlm from the Statistics and Machine Learning Toolbox, but that is mainly due to familiarity, I suppose.

Could it be poor initial guesses for the parameters? For example, if you start with B > 0, I can imagine a fitting routine might struggle.

Can you upload the data here?

[John Terry]

Thanks, i've tampered with the starting parameters and managed to get satisfying result.

Answer 2

[Sulaymon Eshkabilov]

There are a couple of other fit model fcns of the toolbox that can be employed with your exercise. They are nlinfit() and nlintool()

[the cyclist]

I would strongly recommend fitnlm over using nlinfit directly. fitnlm calls nlinfit under the hood, but the resulting model object you get is much easier in general to work with for next steps.